中文摘要：

为探讨利用近地高光谱和遥感影像数据结合预测土壤含水量的可行方法，以黄河三角洲垦利县为研究区，采用中心波长反射率和波段平均反射率两种拟合方法，利用室外实测高光谱窄波段反射率数据模拟LandSat8卫星宽波段反射率，进而通过组合，选取敏感光谱参量，应用多元逐步线性回归方法分别建立土壤含水量高光谱单一形式波段组合与多形式波段组合估测模型，并选取最优估测模型。采用线性混合像元分解处理遥感影像，同时采用比值均值订正方法对遥感影像反射率进行订正，在此基础上，将模型应用到经过订正的LandSat8卫星影像，实现了对研究区土壤含水量的遥感反演。结果表明，最佳模型是基于波段平均反射率拟合方法建立的多形式波段组合估测模型。从反演结果看较为符合研究区土壤含水量的实际状况。

英文摘要：

Acquisition of the information of soil moisture regime is one of the hotspots in current researches. It is not an easy job to achieve inversion of regional soil moisture content just by depending on soil water estimation models established solely on near-ground hyper-spectra. The study is to explore feasible ways to forecast soil moisture contents by combining the use of narrow-band hyper-spectra and wide-band muhi- spectral remote sensing images. Field surveys were conducted and soil samples collected during April 28 to April 30, 2014 in Kenli County, the research area in the Yellow River Delta. Soil water contents were measured in lab using the soil samples and oven-drying method; soil spectra of undisturbed soil samples collected from fields were determined under natural light outdoors with an American ASD Fieldspec4 spectrometer; and the first 7 bands of the OIL sensor were selected and used to collate the LandSat8 remote sensing images of May 1, 2014 for atmospheric radiation correction, geometric precision correction, clipping and other processing. And further on, based on the hyper-spectral narrow-band reflectances measured outdoors LandSat8 wide- band reflectances were simulated with two fitting methods, center wavelength reflectance and band average reflectance methods; by means of band combination in four modes, i.e., ratio, difference, sum dividing reduction, and reduction dividing sum, with sensitive spectral parameters selected according to correlativity; then hyper-spectral single-form band combination and multi-form band combination soil moisture estimation models were established with the multiple stepwise linear regression analysis method, and then screened with the two fitting methods for the best model. Soil information in the remote sensing images was obtained using the linear mixed pixel decomposition method after excluding the vegetation information; the soil information was compared with the measured hyper-spectral reflectance and remote sensing image reflectances were corrected with the r